Skip to main content

    Image to Text (OCR)

    Extract text from images using AI-powered OCR. No image uploads; the OCR runs entirely in your browser. Supports 18+ languages.

    Free to use. Runs in your browser.

    Upload any image and our browser-based OCR engine will extract all readable text instantly.

    Powered by Tesseract.js WASM. Your images never leave your device.

    Drop an image here or click to upload

    PNG, JPG, WebP, BMP, GIF, Max 20 MB

    Methodology and sources

    Formula or method

    Uses Tesseract.js 5.1.1, a WebAssembly build of the Tesseract OCR engine (Apache 2.0). On first use, the engine fetches a language-specific .traineddata model (2-15 MB) from a CDN; subsequent runs use the browser-cached copy. The selected image is analysed entirely in-browser via a Web Worker. The recognition pipeline segments the image into character regions, compares each region against the loaded language model, and returns the best-match character sequence as plain text.

    Basis and assumptions

    • Accuracy figures in the reference table are best-case estimates for each input category under good conditions. Actual accuracy varies with image resolution, lighting, contrast, font, and language.
    • The headline 95-99% accuracy range applies to high-resolution scans and screenshots of standard printed fonts. It does not apply to handwriting, low-resolution photographs, or complex backgrounds.
    • Handwriting recognition is limited. Neat block printing may reach 50-70%; joined cursive is not reliably recognised (20-40%). Tesseract is trained primarily on printed text.
    • The tool processes one image at a time. PDFs must be converted to images before use.
    • Language model files are version 4.0.0 traineddata, fetched from cdn.iforgeapps.com on first use and cached by the browser.

    What this tool does not decide

    • Whether extracted text is correct. Always proofread OCR output before using it in documents, legal submissions, or medical records.
    • Handwriting interpretation for legal or medical purposes. Use a specialist transcription service for documents where accuracy is critical.
    • PDF text extraction. For text-layer PDFs, copy-paste directly from the PDF rather than running OCR on a screenshot.

    Sources

    Last checked: 2026-06-11

    How OCR Works

    OCR (Optical Character Recognition) analyses pixel patterns in an image to identify letters, numbers, and symbols. The software breaks the image into small regions, compares each against a trained model of known characters, and outputs the best match as editable digital text.

    This tool uses Tesseract.js, a WebAssembly port of Google's Tesseract OCR engine. The OCR worker runs in your browser after worker, core, and language files load; the selected image is not uploaded for OCR.

    Modern OCR handles printed text very well (95-99% accuracy with clean images). Handwriting recognition is improving but less reliable. For best results, use clear, high-contrast images with straight, well-lit text.

    What to Expect: Accuracy by Input Type

    Input TypeTypical AccuracyTips
    Printed document (scan)97-99%High-res scan, clean background
    Screenshot with text95-99%Use original resolution, no scaling
    Phone photo of document85-95%Shoot straight-on, good lighting
    Receipt or label80-90%Small text is harder, get close
    Text on complex background60-80%Crop to just the text area first
    Handwriting (neat print)50-70%Block capitals work best
    Cursive handwriting20-40%Not reliable, type it manually

    Getting Better Results

    Do

    Use 300+ DPI images. Higher resolution gives the engine more pixels to work with per character. Low-resolution screenshots of small text are the #1 cause of bad results.

    Do

    Crop to just the text. Remove borders, images, and decorative elements. The less noise the engine has to filter, the better it focuses on the actual characters.

    Do

    Select the right language. Each language has its own character model. English is loaded by default, but switching to the correct language before extraction dramatically improves accuracy for non-English text.

    Avoid

    Watermarked or overlaid images. Text over photos, watermarks across documents, and coloured backgrounds behind text all confuse the recognition engine. Clean, high-contrast input gets clean output.

    Avoid

    Heavily compressed JPGs. JPEG compression artefacts blur the edges of characters, especially at small sizes. If possible, use PNG or a high-quality JPG.

    What to Do After Extraction

    • Count words to check the extracted text length and reading time.
    • Fix the case if the OCR misidentified capitalisation throughout the text.
    • Clean whitespace to remove extra spaces and blank lines that OCR often introduces.
    • Correct OCR errors manually or in your text editor, like "l" misread as "1" or "0" misread as "O".

    Related Tools

    PDF to Word

    Convert PDFs to editable documents

    Word Counter

    Count words in extracted text

    Case Converter

    Fix case after extraction

    Image Cropper

    Crop to text area for better results

    Image Resizer

    Upscale low-res images before OCR

    QR Code Generator

    Turn extracted text into a QR code

    How to use this tool

    1

    Upload an image or paste from clipboard, drag-and-drop works too.

    2

    Select the language of the text in the image for best accuracy.

    3

    Click 'Extract Text' and copy the result to your clipboard.

    Common uses

    • Extract text from scanned documents and PDFs
    • Digitise printed notes, receipts, and invoices
    • Copy text from screenshots and images
    • Convert handwritten notes to digital text
    • Translate text in foreign-language images
    • Extract data from charts and infographics

    Share this tool

    Frequently Asked Questions

    How does Image to Text (OCR) work?
    The tool uses Tesseract.js, a WebAssembly port of the Tesseract OCR engine. It analyses pixel patterns in your image to detect and extract text characters. The OCR runs entirely in your browser. On first use, the engine downloads language model files (2-15 MB) from a CDN; after that the worker and language data are cached and no further network requests are made.
    What image formats are supported?
    PNG, JPG, WebP, BMP, and GIF images up to 20 MB. For best results, use high-resolution images (300 DPI+) with clear, well-contrasted text.
    Which languages are supported?
    18 languages including English, Spanish, French, German, Arabic, Chinese (Simplified and Traditional), Japanese, Korean, Hindi, Russian, Turkish, Polish, Dutch, Swedish, and Vietnamese.
    Is my data private?
    Your images never leave your device. The OCR engine runs entirely inside your browser using WebAssembly, so no image data is ever uploaded or processed on a server. On first use, the engine downloads the OCR language model files from a CDN (the same way a font or library loads); those files contain no personal data. Subsequent runs use the browser's cached copy.
    Why is the first run slow?
    The first extraction downloads the language model (2-15 MB depending on language) and initialises the WASM engine. Subsequent extractions reuse the cached model and are much faster.
    How can I improve OCR accuracy?
    Use high-resolution images with good contrast. Straighten any tilted text, crop to the text area, and select the correct language. Dark text on a light background works best.
    Can I extract text from handwriting?
    Tesseract works best with printed text. It can handle neat, consistent handwriting with moderate accuracy, but cursive or messy handwriting will produce unreliable results. For handwriting, consider a specialised AI tool.
    Does it work with scanned PDFs?
    Not directly, this tool processes images. To extract text from a scanned PDF, take a screenshot of each page or convert the PDF to images first, then use this tool on each image.
    Can I paste images from my clipboard?
    Yes. Click 'Paste from Clipboard' to paste a screenshot or copied image directly. This is especially handy for extracting text from screenshots without saving them as files first.
    What about text in tables or columns?
    Tesseract reads text left-to-right, top-to-bottom. Multi-column layouts and complex tables may produce jumbled output. For best results with tables, crop each column separately.
    How accurate is the text extraction?
    With clear, high-resolution images of printed text, accuracy is typically 95-99%. Lower resolution, unusual fonts, poor lighting, or complex backgrounds reduce accuracy. Always proofread the output.
    Can I extract text from multiple images?
    Currently, the tool processes one image at a time. Extract text from each image individually, then combine the results. You can copy the output to clipboard after each extraction.

    Results are for general informational purposes only and should be checked before use. They are not professional advice. See our Disclaimer and Terms of Service.